1 Department of Otolaryngology Head and Neck Surgery, Medical. 4 School of Nursing, Vanderbilt University Medical Center, Nashville,

Transcription

1 Clinical Techniques and Technology Implementation of Image-Guided Cochlear Implant Programming at a Distant Site Otolaryngology Head and Neck Surgery 2017, Vol. 156(5) Ó American Academy of Otolaryngology Head and Neck Surgery Foundation 2017 Reprints and permission: sagepub.com/journalspermissions.nav DOI: / Theodore R. McRackan, MD 1*, Jack H. Noble 2*, Eric P. Wilkinson 3, Dawna Mills 3, Mary S. Dietrich 4, Benoit M. Dawant 2, Rene H. Gifford, PhD 5, and Robert F. Labadie 6 Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article. Abstract Our objective was to prospectively evaluate implementation of a new cochlear implant (CI) mapping technique, image-guided cochlear implant programming (IGCIP), at a site distant to the site of development. IGCIP consists of identifying the geometric relationship between CI electrodes and the modiolus and deactivating electrodes that interfere with neighboring electrodes. IGCIP maps for 17 ears of 15 adult CI patients were developed at a central image-processing center, Vanderbilt, and implemented at a distant tertiary care center, House Ear Institute. Before IGCIP and again 4 weeks after, qualitative and quantitative measures were made. While there were no statistically significant groupwise differences detected between baseline and IGCIP qualitative or quantitative measures, 11 of the 17 (64.7%) elected to keep the IGCIP map. Computed tomography (CT) image quality appears to be crucial for successful IGCIP, with 100% of those with high-resolution CT scans keeping their maps compared to 53.8% without. Keywords cochlear implant, cochlear implants, cochlear implantation, image guided surgery, quality of life, cochlear implant programming Received October 24, 2016; revised January 18, 2017; accepted February 16, Cochlear implantation (CI) is standard of care for patients with severe to profound sensorineural hearing loss. While most patients perform well, great variability in outcomes exist. Postoperative programming represents an area where improvements could be realized if the location of electrodes in relationship to the modiolus were known. Image-processing techniques to achieve this have recently been recently introduced by a research group at Vanderbilt and clinically used to program CI recipients via a process known as image-guided CI programing (IGCIP). 1 Clinical implementation at the primary clinical site has demonstrated significant performance improvement using IGCIP, 1,2 and we report herein implementation of the technology at a distant site. Materials and Methods Institutional review board approval from Saint Vincent s Hospital was obtained, and patients who were at least 18 years old and had 6 months of use of a traditional CI map programmed by an expert audiologist were recruited. Post-CI temporal bone computed tomography (CT) scans were obtained with minimum 0.4- to 0.5-mm slices nonoverlapping or 0.6- to 0.8-mm slices with 0.3- to 0.4-mm overlapping cuts in the axial plane. CT scans were de-identified and sent to Vanderbilt for evaluation. If preoperative CTs were available, analysis was done as previously reported. 1 First, a high-resolution statistical shape model of the cochlea and its subcompartments was registered to the patient s CT to accurately identify the borders of the modiolus. 3 Next, the positions of individual CI electrodes were found using graph-based electrode localization techniques. 4,5 Next, the 2 CT images were registered to quantify the position of the electrodes relative to the modiolus. Patients for whom no preoperative CT was available were also enrolled, noting that the image processing is more difficult secondary to the metallic artifact caused by the implant. To overcome this, different processes are used. 6,7 Finally, based on the 1 Department of Otolaryngology Head and Neck Surgery, Medical University of South Carolina, Charleston, South Carolina, USA 2 Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, Tennessee, USA 3 House Ear Clinic, Los Angeles, California, USA 4 School of Nursing, Vanderbilt University Medical Center, Nashville, Tennessee, USA 5 Department of Hearing and Speech Sciences, Vanderbilt University Medical Center, Nashville, Tennessee, USA 6 Department of Otolaryngology Head and Neck Surgery, Vanderbilt University Medical Center, Nashville, Tennessee, USA * These authors have contributed equally to this work. Corresponding Author: Theodore R. McRackan, MD, Department of Otolaryngology Head and Neck Surgery, Medical University of South Carolina, 135 Rutledge Avenue, MSC 550, Office #1131 Rutledge Tower, Charleston, SC 29425, USA. mcrackan@musc.edu

2 934 Otolaryngology Head and Neck Surgery 156(5) Table 1. Patient and Device Information for All Patients and Separated Based on Those Who Kept and Did Not Keep Their Image-Guided Factor All Patients (N = 17) Kept Map (n = 11) Did Not Keep Map (n = 6) P Value Age, mean (SD), y 64.6 (11.7) 67.7 (8.9) 58.8 (14.7).137 Bilateral, No. (%) 8 (47.1) 7 (63.6) 1 (16.7).088 Right-sided, No. (%) 9 (52.9) 7 (63.6) 2 (33.3).247 Duration of deafness, mean (SD), y a 11.1 (7.8) 11.0 (7.6) 11.4 (9.1).928 Length of CI use prior to experimental map, mean (SD), y 6.8 (5.1) 7.6 (5.8) 5.3 (3.3).327 Had preoperative CT, No. (%) 4 (23.5) 4 (36.4) 0 (0.0).091 Device, No. (%) Advanced Bionics device 3 (17.6) 1 (5.8) 2 (11.8).210 Cochlear device 14 (82.4) 10 (58.8) 4 (23.5) No. of electrodes deactivated, mean (SD) 9.0 (3.3) 9.1 (3.8) 8.8 (2.6).885 Abbreviations: CI, cochlear implant; CT, computed tomography; SD, standard deviation. a All patients, n = 16; kept map, n = 11; did not keep map, n = 5. Table 2. Speech Testing Outcomes for All Patients and Separated Based on Those Who Kept and Did Not Keep Their Image-Guided Median (IQR) Test No. When Not All Patients Pre, % Post, % Change, % P Value CNC words Did not keep map 51.0 (18, 85) 48.0 (7, 83) 3.0 ( 9, 7).674 Kept map 36.0 (24, 50) 38.0 (20, 56) 2.0 ( 12, 8) All 36.0 (23, 64) 38.0 (18, 61) 2.0 ( 8, 7) CNC phonemes Did not keep map 70.0 (40, 93) 67.0 (15, 93) 2.0 ( 18, 6).178 Kept map 60.0 (50, 68) 64.0 (47, 69) 3.0 ( 3, 8) All 60.0 (48, 77) 64.0 (40, 79) 1.0 ( 3, 8) AzBio quiet Did not keep map 54.5 (13, 98) 54.5 (16, 95) 1.0 ( 5, 3).765 Kept map 38.0 (21, 58) 43.0 (27, 58) 5.0 ( 7, 9) All 38.0 (19, 77) 43.0 (24, 79) 0.0 ( 5, 9) AzBio 110 Did not keep map 24.0 (3, 83) 24.0 (3, 81) 0.5 ( 5, 2).348 Kept map 7.0 (0, 15) 9.0 (7, 28) 2.0 (0, 8) All 7.0 (2, 32) 9.0 (5, 30) 1.0 ( 1, 6) AzBio 15 Did not keep map (3) 40.0 (19, ) 31.0 (13, ) 6.0 ( 36, ) a Kept map (1) 16.0 (16, 16) 17.0 (17, 17) 1.0 (1, 1) All (4) 29.5 (16, 61) 24.0 (14, 43) 5.5 ( 29, 6) BKB-SIN Did not keep map 13.5 (7, 22) 14.3 (7, 24) 0 ( 2, 4).641 Kept map 17.0 (15, 21) 16.5 (14, 23) 0 ( 2, 2) All 17.0 (13, 21) 16.5 (11, 23) 0.5 ( 2, 3) QSMD average Did not keep map (6) 65.8 (37, 90) 44.5 (32, 55) 21.3 ( 42, 5).038 Kept map (9) 45.5 (34, 61) 48.2 (38, 75) 3.0 ( 5, 9) All (15) 53.6 (37, 75) 48.7 (35, 69) 7.9 ( 21, 5) Abbreviations: BKB-SIN, Bamford-Kowal-Bench Speech-in-Noise; CNC, consonant-vowel nucleus-consonant; IQR, interquartile range; QSMD, quick spectral modulation detection. a Due to the small sample size, no statistical analysis was performed. position of the electrodes relative to the modiolus, electrodes that are predicted to have a high probability of interfering with neighboring electrodes were deactivated from the patient s map. 1 Each patient underwent a hearing and speech recognitions test (Minimum Speech Test Battery and Bamford- Kowal-Bench Speech-in-Noise [BKB-SIN]), qualitative testing (Abbreviated Profile of Hearing Aid Benefit [APHAB] and Speech, Spatial and Qualities of Hearing Scale [SSQ]), and a quick spectralˇmodulation detection (QSMD) task, a non-speech-based hearing performance metric that provides a psychoacoustic estimate of spectral resolution. After

3 McRackan et al 935 Table 3. Qualitative Outcomes for All Patients and Separated Based on Those Who Kept and Did Not Keep Their Image-Guided Mean (SD) Test No. When Not All Patients Pre, % Post, % Change, % P Value APHAB global Did not keep map (5) 44.1 (18.8) 42.8 (11.4) 1.3 (16.6).709 Kept map (11) 48.1 (21.2) 43.7 (21.4) 4.4 (8.3) All (16) 46.8 (19.9) 43.4 (18.4) 3.5 (11.0) APHAB EC Did not keep map (5) 26.2 (20.0) 25.1 (13.2) 1.1 (12.6).797 Kept map (11) 27.2 (27.4) 24.4 (22.8) 2.8 (12.9) All (16) 26.9 (24.4) 24.6 (19.8) 2.3 (12.4) APHAB RV Did not keep map (5) 58.4 (26.1) 55.2 (8.3) 3.2 (24.0).865 Kept map (11) 58.7 (21.1) 53.9 (23.5) 4.7 (13.2) All (16) 58.6 (22.3) 54.3 (19.7) 4.3 (16.4) APHAB BN Did not keep map (5) 47.7 (20.4) 48.1 (14.1) 0.5 (16.0).594 Kept map (11) 58.4 (20.8) 52.8 (22.5) 5.6 (11.8) All (16) 55.1 (20.2) 51.3 (19.9) 3.7 (13.1) APHAB AV Did not keep map (5) 30.5 (20.6) 30.7 (24.1) 0.2 (13.5).562 Kept map (11) 35.4 (21.5) 30.7 (20.2) 4.7 (12.5) All (16) 33.9 (20.5) 30.7 (20.7) 3.2 (12.6) SSQ speech Did not keep map 4.8 (1.9) 4.1 (2.0) 0.7 (0.9).074 Kept map 4.6 (2.1) 4.9 (2.0) 0.2 (0.9) All 4.7 (2.0) 4.6 (2.0) 0.1 (1.0) SSQ spatial Did not keep map 4.3 (1.7) 3.9 (1.9) 0.4 (1.0).078 Kept map 4.0 (2.3) 4.7 (2.7) 0.7 (1.2) All 4.1 (2.0) 4.4 (2.4) 0.3 (1.2) SSQ qualities Did not keep map 5.8 (2.1) 5.2 (1.4) 0.6 (1.3).077 Kept map 5.4 (2.5) 5.9 (2.7) 0.5 (0.9) All 5.6 (2.3) 5.6 (2.3) 0.1 (1.2) SSQ total Did not keep map 4.9 (1.8) 4.5 (1.5) 0.5 (0.4).007 Kept map 4.8 (2.2) 5.2 (2.3) 0.5 (0.6) All 4.8 (2.0) 4.9 (2.1) 0.1 (0.7) APHAB, Abbreviated Profile of Hearing Aid Benefit; AV, aversiveness; BN, background noise; EC, ease of communication; RV, reverberation; SSQ, Speech, Spatial and Qualities of Hearing Scale. baseline testing, the participant s CI was reprogrammed according to the IGCIP strategy and returned 4 weeks later for repeat testing. Participants were provided the IGCIP map only to ensure compliance. Statistical analyses were done using SPSS (version 23; SPSS, Inc, an IBM Company, Chicago, Illinois) with appropriate tests (eg, logistic regression to assess whether the amount of change in speech increased or decreased the likelihood of keeping the map after controlling for the respective pre-igcip value). Results Seventeen ears in 15 postlingually deafened CI users were enrolled (see Table 1). On average, patients had 6.8 years of CI use prior to enrollment. With IGCIP, patients with Cochlear devices (Cochlear, Englewood, Colorado, USA) had, on average, 12.3 electrodes active, and those with Advanced Bionics (Advanced Bionics LLC, Valencia, California, USA) had 10. At the end of the 4-week trial, 11 patients (64.7%) chose to keep their IGCIP map. No patient factors were statistically associated with keeping the IGCIP map (all P..05). Figure 1 displays the individual patient data for the qualitative and quantitative metrics (note APHAB and BKB-SIN axes are reversed because decreased score denotes improvement). Hearing outcomes for the group prior to and after implementing the study map are displayed in Table 2. As a group, there were no statistically significant changes in hearing outcomes or qualitative measures when comparing IGCIP performance to baseline (all P..05) (Table 3). All patients who had a preoperative CT kept their IGCIP. Discussion This study represents the first reported use of IGCIP at a distant site. Patients were recruited, underwent testing, and had IGCIP maps implemented at House Ear Institute (HEI), while the IGCIP plan was made at Vanderbilt. In total, 64.7% of patients elected to keep their IGCIP map despite, on average, no statistically significant detectable improvement in patient word recognition. This is somewhat lower than the 78% to 82% previously reported by the home

4 936 Otolaryngology Head and Neck Surgery 156(5) Figure 1. Individual subject data before and after image-guided cochlear implant programming (IGCIP) (subject ID is consistent throughout). Blue data points represent those who kept IGCIP and gray are those who did not. APHAB, Abbreviated Profile of Hearing Aid Benefit; CNC, consonant-vowel nucleus-consonant; SSQ, Speech, Spatial and Qualities of Hearing Scale. institution.1,2 We hypothesize that this may be due to one of the following 3 reasons. First, with any new technology, subtleties arise during implementation that may affect outcomes. Second, the sample size at HEI was an order of magnitude smaller than at Vanderbilt. With enrollment of more patients, it is possible that similar outcomes will be seen. Third, and most interesting from a scientific standpoint, is the impact of CT scan characteristics to the sensitivity of the IGCIP image analysis as this was the first time that CT images were obtained at various ambulatory sites. Based on the findings of this study, we have undertaken phantom studies and identified that slice thickness 0.5 mm (12 of 17 images acquired in this study) and those not reconstructed with extended Hounsfield unit ranges (all images acquired in this study) have less accurate IGCIP processing. In addition, prior work has suggested that the use of post-ci CTs alone should be adequate for determining intracochlear CI electrode position,7 but the sensitivity of this approach to postoperative CT characteristics has yet to be quantified. In the current study, it was found that all patients with pre-ci CTs kept their study maps. Conclusion IGCIP is a relatively new technology that can be successfully implemented at outside institutions. The data presented support continued efforts at more widespread clinical implementation. High-resolution CT scanning with extended Hounsfield unit ranges appears to be critical to IGCIP.

5 McRackan et al 937 Author Contributions Theodore McRackan, acquisition, analysis, interpretation of data and manuscript preparation, final approval, accountability for all aspects of the work; Jack Noble, acquisition, analysis, interpretation of data, manuscript preparation, final approval, accountability for all aspects of the work; Eric Wilkinson, manuscript preparation and analysis of data, final approval, accountability for all aspects of the work; Dawna Mills, data acquisition, drafting and revising manuscript; Mary S. Dietrich, data analysis, drafting and revising manuscript; Benoit Dawant, analysis, interpretation of all aspects of the work; Rene Gifford, analysis, interpretation of all aspects of the work; Robert Labadie, analysis, interpretation of all aspects of the work. Disclosures Competing interests: Eric Wilkinson, grant support from Cochlear Americas and MED-EL; consultant for Nanotron Corporation. Dawna Mills, Advanced Bionics advisory board. Benoit Dawant, cofounder and equity holder for Neurotargeting, LLC. Rene Gifford, Audiology Advisory Board for Cochlear Americas, Advanced Bionics, and Frequency Therapeutics; past consultant for Medel. Robert Labadie, consultant for Advanced Bionics, Cochlear, Medtronic, and Ototronix. Sponsorships: Funded by below sources. No sponsor. Funding source: This research was made possible by funding from the National Institutes of Health (NIH) grants R01DC014037, R21DC012620, and R01DC from the National Institute of Deafness and Other Communications Disorders; a K12 award through the South Carolina Clinical & Translational Research (SCTR) Institute, with an academic home at the Medical University of South Carolina; NIH/National Center for Advancing Translational Sciences grant UL1TR001450; and a grant from the Doris Duke Charitable Foundation. References 1. Noble JH, Labadie RF, Gifford RH, Dawant BM. Image-guidance enables new methods for customizing cochlear implant stimulation strategies. IEEE Trans Neural Syst Rehabil Eng. 2013;21: Noble JH, Gifford RH, Hedley-Williams AJ, Dawant BM, Labadie RF. Clinical evaluation of an image-guided cochlear implant programming strategy. Audiol Neurootol. 2014;19: Noble JH, Gifford RH, Labadie RF, Dawant BM. Statistical shape model segmentation and frequency mapping of cochlear implant stimulation targets in CT. Med Image Comput Comput Assist Interv. 2012;15(pt 2): Zhao Y, Dawant BM, Labadie RF, Noble JH. Automatic localization of cochlear implant electrodes in CT. Med Image Comput Comput Assist Interv. 2014;17(pt 1): Noble JH, Dawant BM. Automatic graph-based localization of cochlear implant electrodes in CT. Med Image Comput Comput Assist Interv. 2015;9350: Reda FA, Noble JH, Labadie RF, Dawant BM. An artifactrobust technique for automatically segmenting the labyrinth in post-cochlear-implantation CT. Proc SPIE Conf Med Imaging. 2014;9034:9034V. 7. Reda FA, McRackan TR, Labadie RF, Dawant BM, Noble JH. Automatic segmentation of intra-cochlear anatomy in postimplantation CT of unilateral cochlear implant recipients. Med Image Anal. 2014;18: